[ewg] OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time (12pm EST) - Minutes

Thu Feb 1 11:59:10 PST 2018

And 2663 has been fixed in rdma-core-16.2 stable release.  I've asked Vlad
to pull in that release.

Steve.

From: ewg [mailto:ewg-bounces at lists.openfabrics.org] On Behalf Of Davis,
Arlin R
Sent: Thursday, February 01, 2018 1:25 PM
To: Kalderon, Michal; 'ewg at lists.openfabrics.org'
Cc: Srikakulam, Venkata
Subject: Re: [ewg] OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US
Pacific Time (12pm EST) - Minutes

Michal, thanks for the update.

Thanks Michal for the update. This is what we have so far: 

 <http://bugs.openfabrics.org/show_bug.cgi?id=2662> Bug 2662 Chelsio: Cannot
set mtu greater than 1500 on SLES12Sp3 - Fixed/Closed, Thanks!
 <http://bugs.openfabrics.org/show_bug.cgi?id=2668> Bug 2668 Broadcom:
Cannot change MTU to greater than default - No ETA on Fix
 <http://bugs.openfabrics.org/show_bug.cgi?id=2669> Bug 2669 Cavium: Cannot
set mtu greater than 1500 on SLES12Sp3 - ETA for fix, next week 

We will need fixes before moving to RC1. 

-arlin

From: Kalderon, Michal [mailto:Michal.Kalderon at cavium.com] 
Sent: Thursday, February 01, 2018 7:08 AM
To: Davis, Arlin R <arlin.r.davis at intel.com>; Schmidt, William R
<william.r.schmidt at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Cc: Srikakulam, Venkata <Venkata.Srikakulam at cavium.com>; Nikolova, Tatyana E
<tatyana.e.nikolova at intel.com>; Pradeep Kankipati
<pradeep.kankipati at broadcom.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

Opened bugzilla: http://bugs.openfabrics.org/show_bug.cgi?id=2669

Will provide a patch to qedr next week. 

Thanks, 
Michal

From: Davis, Arlin R [mailto:arlin.r.davis at intel.com] 
Sent: Wednesday, January 31, 2018 8:56 PM
To: Kalderon, Michal <Michal.Kalderon at cavium.com>; Schmidt, William R
<william.r.schmidt at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Cc: Srikakulam, Venkata <Venkata.Srikakulam at cavium.com>; Nikolova, Tatyana E
<tatyana.e.nikolova at intel.com>; Pradeep Kankipati
<pradeep.kankipati at broadcom.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

True, thanks for catching this. Please open a qedr bug so we can track as
critical/blocking.

Tatyana and Pradeep, do Intel and/or Broadcom drivers need similar changes? 

Anyone else?

From: Kalderon, Michal [mailto:Michal.Kalderon at cavium.com] 
Sent: Wednesday, January 31, 2018 10:29 AM
To: Davis, Arlin R <arlin.r.davis at intel.com>; Schmidt, William R
<william.r.schmidt at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Cc: Srikakulam, Venkata <Venkata.Srikakulam at cavium.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

We assumed this is a generic issue since it happened on several adapters, I
see now with the bug resolve that Steve made changes specific to cxgb3
We need to make similar changes in qede.

Thanks,
Michal

From: Davis, Arlin R [mailto:arlin.r.davis at intel.com] 
Sent: Wednesday, January 31, 2018 8:10 PM
To: Kalderon, Michal <Michal.Kalderon at cavium.com>; Schmidt, William R
<william.r.schmidt at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Cc: Srikakulam, Venkata <Venkata.Srikakulam at cavium.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

Are you using the latest daily builds? Steve Wise reported this and
forwarded patches to Vlad on Jan 19th. 

Fix went into build:
http://downloads.openfabrics.org/OFED/ofed-4.8-2-daily/OFED-4.8-2-20180122-1
411.tgz

Bug: http://bugs.openfabrics.org/show_bug.cgi?id=2662

Please let us know if you still have issues with latest builds.

-arlin

From: Kalderon, Michal [mailto:Michal.Kalderon at cavium.com] 
Sent: Wednesday, January 31, 2018 9:15 AM
To: Schmidt, William R <william.r.schmidt at intel.com>; Davis, Arlin R
<arlin.r.davis at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Cc: Srikakulam, Venkata <Venkata.Srikakulam at cavium.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

Hi

We're seeing an issue on SLES12SP3 with modifying mtu. 

linux-p4eo:~ # ifconfig eth7 mtu 9000
SIOCSIFMTU: Invalid argument
Dmesg: eth7: Invalid MTU 9000 requested, hw max 1500

We've seen this with other vendor devices as well. 
Chelsio reported an issue in the past regarding MTU change,
But I didn't see any related bugs open or discussions. 

Has this been discussed since? 
Attaching relevant email

Thanks,
Michal

From: ewg [mailto:ewg-bounces at lists.openfabrics.org] On Behalf Of Schmidt,
William R
Sent: Monday, January 29, 2018 10:15 PM
To: Davis, Arlin R <arlin.r.davis at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Subject: Re: [ewg] OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US
Pacific Time (12pm EST) - Minutes

Yes please. The omitted commits are listed in the OFED bugs.

From: Davis, Arlin R 
Sent: Monday, January 29, 2018 2:06 PM
To: Schmidt, William R <william.r.schmidt at intel.com>;
'ewg at lists.openfabrics.org' <ewg at lists.openfabrics.org>
Cc: vlad at dev.mellanox.co.il; Woodruff, Robert J
<robert.j.woodruff at intel.com>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

Bill, thanks for the update. Do you need Vlad's help getting these fixes
into OFED 4.8 compat-rdma?

From: Schmidt, William R 
Sent: Monday, January 29, 2018 10:37 AM
To: Davis, Arlin R <arlin.r.davis at intel.com>; 'ewg at lists.openfabrics.org'
<ewg at lists.openfabrics.org>
Subject: RE: OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific Time
(12pm EST) - Minutes

>>2. New bugs:  True Scale qib bonding issue on RH 7.4, SLES 12.1 and 12.2 -
Need update from Bill Schmidt (Intel)

Bug <http://bugs.openfabrics.org/show_bug.cgi?id=2664>  2664 - Bonding
doesn't work on RHEL 7.4
Bonding driver form RHEL 7.4 has integrated commit:
<https://github.com/torvalds/linux/commit/b5bf0f5b16b9c316c34df9f31d4be8729e
b86845>
https://github.com/torvalds/linux/commit/b5bf0f5b16b9c316c34df9f31d4be8729eb
86845
This requires ipoib driver to return correct speed and duplex mode.  Commit
adding this feature on ipoib driver:

<https://github.com/torvalds/linux/commit/0d7e2d2166f6b0b7d1959ca858052a15fe
b574cc>
https://github.com/torvalds/linux/commit/0d7e2d2166f6b0b7d1959ca858052a15feb
574cc was added in 4.12 kernel, so it is missing from OFED 4.8 compat-rdma.
In consequence Bonding driver cannot retrieve required data and fails.

 <http://bugs.openfabrics.org/show_bug.cgi?id=2665> Bug 2665 - Bonding
causes kernel panic on SLES 12.1 and SLES 12.2
Bonding driver from SLES 12.1/12.2 is missing kernel panic fix commit:
<https://github.com/torvalds/linux/commit/1533e77315220dc1d5ec3bd6d9fe32e2aa
0a74c0>
https://github.com/torvalds/linux/commit/1533e77315220dc1d5ec3bd6d9fe32e2aa0
a74c0
added in linux kernel 4.8. This makes it incompatible with ipoib drivers
from OFED 4.8 compat-rdma.

From: ewg [mailto:ewg-bounces at lists.openfabrics.org] On Behalf Of Davis,
Arlin R
Sent: Monday, January 29, 2018 11:57 AM
To: 'ewg at lists.openfabrics.org' <ewg at lists.openfabrics.org>
Subject: [ewg] OFA EWG Meeting: Monday, Jan 29, 2017, 09:00 AM US Pacific
Time (12pm EST) - Minutes

Attendees                      Company     
Pradeep Kankipati         Broadcom      
Steve Wise                     Chelsio
Robert Woodruff            Intel
Arlin Davis                      Intel                                 
Vladimir Sokolovsky       Mellanox                            
Ariel Elior                       Cavium   
Michal Kalderon             Cavium   
Michael Rice                       HPE

Minutes:   

*         Opens - none

*	OFED 4.8-2:
<http://downloads.openfabrics.org/OFED/ofed-4.8-2-daily/OFED-4.8-2-20180124-
0818.tgz>
http://downloads.openfabrics.org/OFED/ofed-4.8-2-daily/OFED-4.8-2-20180124-0
818.tgz

Status:  RH7.4 and SLES12.3 backports added. 
                Updated packages:  rdma_core-v16, perftest 4.1-0.2,
libfabric 1.5.3
                Installation changes:  --without-depcheck docs, vmw_pvrdma
moved out of tech preview
             Test Status:  Intel - build RH 7.0, 7.1, 7.2, 7.3, 7.4 SLES 12,
12.1, 12.2, and 12.3 - Passed

Known issues to be resolved before RC1:

1. Bug #2663 - (P1) rping fails, iwpmd hitting segfault on SLES12.3 - 
*	Chelsio validation team hits bug, engineering team cannot reproduce.

*	Steve (Chelsio) needs help from Tatyana's (Intel) team to reproduce
and isolate.
2. New bugs:  True Scale qib bonding issue on RH 7.4, SLES 12.1 and 12.2 -
Need update from Bill Schmidt (Intel)

*	OFED 4.8-2 RC1 schedule: (2 blocking bugs) 

Plan is to clean up bugs this week and push hard for RC1 by Friday.
                The GA plan is to go from RC1 to GA, 1-2 week RC1
validation, and Feb 16th for a GA target.
Board approved OFED 4.8-2 so we can move to GA as soon as EWG is ready.

Regards,

Arlin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ewg/attachments/20180201/15246144/attachment.html>